Filip Novokmet, Thomas Piketty, Li Yang, Gabriel Zucman · AEA Papers and Proceedings 108: 109-113

From Communism to Capitalism: Private versus Public Property and Inequality in China and Russia

16 problems found: 1 major · 6 moderate · 9 minor

What the paper claims

The transition produced a massive shift from public to private property: private wealth rose much faster than national income in both countries, and Russian public wealth collapsed with voucher privatization.
Housing drove the rise of private wealth; Russian households hold little recorded financial wealth but large unrecorded offshore wealth; public ownership is no longer dominant but remains far larger than in Western countries.
Income inequality rose in both countries, but much faster and to higher top-end levels in Russia; China's rise was more gradual.
Growth incidence: China's much faster growth substantially lifted the bottom; Russia's post-communist growth was captured by the top.
Forbes billionaire data show a much greater concentration of wealth among billionaires in Russia than in China and Western countries.
The divergent paths partly reflect the different privatization strategies (gradual in China, shock therapy/vouchers in Russia); Russia's rise also exceeds that of more gradually privatizing Eastern European countries; rising inequality is not inevitable and policies, institutions and ideology matter.

Every problem found

1majorlikely · previously known · speculative assumption

Russian tax records list only people's gross takings, such as total proceeds from selling shares or business turnover, so the authors had to guess how much of that is actual income. They assumed that 30% of the takings at the top are income, more than the tax records themselves imply, even though top filers are largely business owners and share sellers with large costs. With more realistic assumptions the richest 1% of Russians get roughly 16-19% of national income rather than 20%. Russia's lead over China shrinks by a quarter to two-thirds, but Russia stays ahead.

Technical detail and evidence

Why it matters, in detail. The fiscal correction multiplies RLMS incomes above P90 by piecewise-linear factors anchored at receipt thresholds x (1 - d). The factor at the 10m-ruble knot (p ~ 0.996, ratio 2.23 in 2015) scales the entire P90-P99.6 range, so the top-1% share is close to linear in (1 - d) there. With d = 0.70 the 2015 fiscal top 1% is 17.93% (pretax 20.24%). With the observed d = 0.739 it is 16.29% (pretax ~18.8%), and with d = 0.80 at the knots 13.46% (pretax ~16.2%). Against China's 13.92%, the 2015 gap moves from 6.3pp to ~4.8pp (-23%), to 4.6pp in the current WID vintage (-27%), and to ~2.3pp at d = 0.80 (-63%). Russia's top 10% would fall from 45.5% to ~40-43%, i.e. close to China's 41.4%. The evidence (observed aggregate, bracket composition, later vintage, unit mismatch) points to overstatement. Declaration-rate uncertainty is two-sided (NPZ variants 15.8-24.6% fiscal).

Evidence. NPZ (2017) Appendix B pp. 18-19; NPZ2017AppendixB.xlsx TB13 (B5:B9 declaration profile; O5:V9 benchmark deductions 0.80/0.80/0.80/0.75/0.75/0.70/0.70/0.70 for 2008-15; W5:AA9 variant multipliers, all equal to 1 at the 10m row; A20:H20 observed aggregate deduction 0.7749 ... 0.7393), TB10 rows 53-57, TB11 col J (entrepreneur shares). Adjudicator re-implementation of do_gpinter_RussiaRLMS.do (scripts/russia_taxcorr.py) reproduces all nine published 2015 variants of the fiscal top-1% share to four decimals (benchmark 17.93%; declaration variants 15.82-24.60%; deduction variants 16.55-19.27%; Appendix B Fig. B41 chart cache). Alternatives (2015, fiscal top 1%): observed aggregate 0.739 -> 16.29%; 0.75 -> 15.82%; 0.80 at the three knots -> 13.46%; 0.85 -> 10.82% (identical to Auditor B's figures). Pretax equivalents are roughly 18.8% and 16.2% (Auditor B's copula approximation; benchmark 20.24%). The 'floor at one' in the code never binds: correction ratios are 2.05-36.3 in every year 2008-2015 (work/russia_taxcorr_years.log). The current WID.world vintage puts Russia's 2015 top 1% at 18.6% and China's at 14.0% (work/wid_current.log). The sibling adjudication of NPZ (2018) reaches the same conclusion (major/likely) and cites 2023 tax counts consistent with ~16-19%.

Why this severity. Russia stays above China in every defensible variant, so the direction survives. But the gap the paper calls 'markedly higher' (6.3pp in 2015) shrinks by about a quarter under the observed deduction rate or the authors' own later WID vintage, and by about two-thirds under a deduction profile consistent with the bracket composition. Russia also falls below the US level that it is said to be 'approaching'. This is an unvalidated assumption that, changed within a defensible range, substantially weakens a core comparative finding.

The authors' response. NPZ (2017, Appendix B pp. 18-20; 2018, p. 216) acknowledge that the tabulations are imperfect, publish declaration- and deduction-rate variants, and argue that in all variants corrected inequality is far above raw survey levels. They also cite the leaked 2004 Moscow tax data (Guriev & Rachinsky 2006), in which tax data raise the survey Gini from ~0.3-0.4 to over 0.6. No reply to Kapeliushnikov (2020) was located.

2moderateconfirmed · newly identified here · methodology

To compare billionaire fortunes with the size of each economy, the authors converted Russia's national income into dollars at the 2016 exchange rate, when the ruble was very weak, instead of at each year's actual rate. That made Russia's economy look about half its real dollar size in 2008-2013 and roughly doubled the apparent weight of its billionaires. Corrected, Russian billionaires still hold much more wealth relative to the economy than in the US, Europe or China, but about twice the US level rather than three times.

Technical detail and evidence

Why it matters, in detail. With a constant-real-exchange-rate denominator, Y_t(USD) = Y_t(real, 2016 $) x P_t(US) understates nominal-dollar NI whenever the RUB real exchange rate was stronger than in 2016. Russia's denominator is understated by 1.42x (2007), 1.68x (2008) and 1.84x (2011-13), relative to NPZ's own Appendix-A nominal-dollar NI. The plotted 2011-13 Russian ratios (36.7-40.3%) fall to 20.9-22.1% and Russia/US falls from ~3.2x to ~1.8x. Corrected, Russia (17-28% in 2010-16) is still roughly 1.7-2x the US, 1.8-2.8x Germany, 2.7-4.3x France and 4-6x China.

Evidence. NPYZ2018Data.xlsx Russia3: CE38 = CJ38*CL38 = $982bn for 2011 and BS38 = 1e6*BR38*BQ38/CE38 = 44.1%. NPZ's own Appendix A Table A33 (col. B = nominal NI in rubles / current RUB-USD rate) gives $1,532bn (2008), $1,803bn (2011), $1,937bn (2013), against Figure-5 denominators of $911bn, $982bn and $1,051bn (ratios 1.68-1.84). World Bank NNI (GNI minus CFC, current US$) gives the same ratio (1.83 in 2011). Keeping the authors' numerators and moving-average windows, the plotted Russian citizen series becomes 2007 36.0 -> 24.4%, 2008 42.5 -> 27.1%, 2011 36.7 -> 20.9%, 2012 40.3 -> 21.9%, 2013 38.4 -> 22.1%, 2014 36.6 -> 25.3%, 2016 28.5 -> 27.8%. Russia/US falls from 3.0-3.5x to 1.7-1.9x in 2011-13. France and Germany are overstated by 1.2-1.4x in 2007-13, and China is understated in 2005-08. The published figure matches the workbook within 0.07pp (digitized), the chart7.xml cache equals Russia3 exactly, and Forbes 2013 numerators match the Forbes list exactly (110 Russians, $427.1bn). Scripts: recompute_fig5.py, digitize_fig5.py, chart_cache_vs_sheet.py.

Why this severity. This is a confirmed error that roughly halves Russia's displayed lead over the West and removes the 2008-2013 'plateau' (the corrected peak is 2016). But Figure 5 is corroborating evidence, the text draws only a qualitative conclusion from it, and Russia remains the highest country in every year after correction. So it affects a secondary result materially and the core finding only modestly.

3moderateconfirmed · newly identified here · misrepresentation of source

The paper says the richest 1% of Russians took more than two-thirds of all income growth between 1989 and 2016, but the authors' own source puts it at 56%. That is still more than half, and the richest 10% took almost all of the growth, so the main point stands. The headline number is nonetheless wrong.

Technical detail and evidence

Why it matters, in detail. For an anonymous fractile with shares s0 and s1 and aggregate growth factor g, the capture share is (g*s1 - s0)/(g - 1). With the authors' own inputs this is 0.564, not above 0.667. The growth-incidence profile (top 10% 98.6%, bottom 50% -15%) is unaffected, so only the quoted magnitude is wrong.

Evidence. Recomputed: capture = (g*s1 - s0)/(g - 1) with g = 23,181/16,432 = 1.4108 (Russia1!B154/B127), s0 = 5.40% (MainData!AC96), s1 = 20.24% (AC122, 2015 shares used for 2016 as in the workbook) gives 56.4%. NPZ (2017) Table 2 and NPZ (2018) report 56% (top 1% income +429%, aggregate +41%). The top 10% captured 98.6%, the middle 40% 16.4% and the bottom 50% -15.0%, whose income fell 20.5%. No window or vintage gives more than two-thirds. The same sentence is in the WID working paper. Script: russia_table1_capture.py.

Why this severity. The headline statistic is overstated by at least 10.3pp (at least 18% relative). The qualitative claim that the bulk of growth went to the top survives, since the top 10% captured almost all of it.

4moderateconfirmed · newly identified here · overclaiming

The paper suggests that Russia's 'shock therapy' privatization caused its inequality explosion while China's gradual approach held inequality down. That may be true, but comparing two very different countries cannot prove it, because they also differed in growth, oil wealth, inflation and politics. The Czech Republic privatized just as quickly with vouchers yet stayed fairly equal, and the authors' longer Russia paper says labour-market changes mattered too.

Technical detail and evidence

Why it matters, in detail. With two treated units and many concurrent differences, the privatization-speed effect is not identified and no share of the 2015 gap can be attributed to it. The like-for-like fiscal comparison narrows the Russia-Poland top-1% gap from 6.8pp (pretax vs fiscal) to 4.5pp, which closes under the F1 alternatives. The rapid Czech voucher privatization with low top shares is a counterexample to speed alone, though compatible with a property-structure channel (the Czech public share of national wealth is China-like).

Evidence. P&P pp. 109, 112-113. NPYZ2018Data.xlsx Russia4: Russia fiscal top 1% 17.9% (2015) vs Poland 13.4% and Czech Republic 9.6% (fiscal); Russia pretax 20.2%. Svejnar & Singer (1994), 'Using vouchers to privatize an economy: the Czech and Slovak case', Economics of Transition 2(1): 43-69, doi:10.1111/j.1468-0351.1994.tb00103.x. NPZ (2018) p. 220 ('We certainly do not mean to suggest ... that the only reason for higher top income shares in Russia is entirely due to different privatization strategies ... the data that we use ... does not allow us to properly disentangle these different factors').

Why this severity. The privatization-strategy explanation is the paper's framing contribution (abstract: 'observe their impacts'), yet it is an untested interpretation; the descriptive divergence it explains is unaffected. Hedges such as 'partly reflects' and 'suggest' limit, but do not remove, the overstatement.

The authors' response. The text is partly hedged ('partly reflects', 'suggest'); NPZ (2018, p. 220) explicitly say privatization is not the only reason and that labour-market forces likely matter.

5moderatelikely · previously known · methodology

There are no Russian tax records for the 1990s, and the authors switch from one household survey to another in 1994. The two surveys tell very different stories: one has the poorer half's share of income falling below 10% in 1996, the other only to about 25%. Inequality certainly jumped after 1991, but how deep the collapse was depends heavily on which survey you believe.

Technical detail and evidence

Why it matters, in detail. The corrected 1996 bottom-50 share (9.6%) tracks raw RLMS (9.9%); the HBS-based alternative is 25.4%. Within HBS the 1991-96 bottom-50 decline is -4.3pp vs -19.4pp in the spliced series. The '>45% in 1996' top-10 figure is likewise an RLMS level (raw 46.0%, corrected 48.3%). Post-1996 the two surveys trend in opposite directions, so the 1996-2007 path also depends on the survey choice and on interpolated Pareto coefficients.

Evidence. NPYZ2018Data.xlsx Russia4 (G-J corrected pretax, L-O HBS, Q-T RLMS). In 1996 the bottom-50 share is 9.6% corrected, 9.9% raw RLMS and 25.4% HBS. The 1991-1996 change in the bottom-50 share is -19.4pp corrected vs -4.3pp within HBS, and the top-10 change is +23.7pp vs +3.6pp. RLMS-HBS top-10 gaps are +10pp (1994), +19pp (1996) and -7pp (2015). NPZ (2018) p. 197 and fn 11-12: RLMS preferred for micro-files and a detailed questionnaire; 'it would be preferable to have access to income tax tabulations prior to 2008. Unfortunately no such data seems to exist'. NPZ Appendix B, TB0 (Pareto coefficients). Kapeliushnikov (2020, HSE WP3/2020/01, pp. 21-22, Fig. 6): structural break 1993/94; the two series 'cannot in principle be combined'.

Why this severity. The 1996 trough and the size of the 1991-96 rise, prominent in the text, are source-dependent by ~10-15pp. The fact of a fast post-1991 rise (also seen in Rosstat), the long-run decline of the bottom 50% (29-30% to 17%) and the tax-anchored 2008-15 levels are not affected.

The authors' response. NPZ (2018, fn 11-12) justify the RLMS (micro-files, detailed income questions; HBS income measured indirectly with opaque tabulations) and call the 1990-2007 profile 'relatively robust', while acknowledging that pre-2008 tax data would be preferable.

6moderatelikely · previously known · speculative assumption

China only has tax records for 2006-2010, so the authors assumed surveys missed top incomes in 1978, at the end of the Mao era, by as much as in the 2000s. That makes China look more unequal at the start and its later rise smaller, while Russia's Soviet starting point got only a small adjustment. Without this assumption the incomes of China's very richest grew as fast as or faster than Russia's. The finding that China's poorer half did far better than Russia's is not affected.

Technical detail and evidence

Why it matters, in detail. Because the tax correction multiplies China's raw 1978 shares by 1.8 (top 1%), 4.8 (top 0.1%), 13 (top 0.01%) and 26 (top 0.001%), China's top-end CAGRs are mechanically depressed. Without the 1978 upgrade they rise by 1.8-10pp/yr and exceed Russia's at the top 0.1% and above. For the top-10%/bottom-50% contrast the transplant is conservative about China's own rise but, in this comparison, flatters the 'gradual China vs abrupt Russia' contrast: the 1978/1980 starting gap in top-10 shares is 6.1pp corrected vs 1.9pp raw.

Evidence. PYZ (2019) AER p. 2477, fn 19. PYZ2017FinalDistributionSeries gperc files (scripts/china_baseline.py). China 1978 top 10/1/0.1/0.001%: raw fiscal 21.85/3.08/0.41/0.01% -> final pretax 27.15/6.35/2.29/0.32%. Russia 1980 (Russia4): HBS/Soviet survey 19.93/2.87% -> final 21.02/3.45%. Table 1 China column reproduced exactly from the final g-percentiles. Removing the 1978 upgrade (2015 unchanged) gives top 1/0.1/0.01/0.001% growth of 10.24/13.82/17.78/20.56%/yr vs published 8.43/9.09/9.83/10.42. Half the upgrade gives 9.19/10.59/11.69/12.39 (arithmetic) or 9.3/11.4/13.7/15.4 (geometric, Auditor A). Raw-to-raw gives 8.34/9.79/11.05/12.03. Russia's figures are 6.24/9.90/13.44/17.01. China's bottom-50 growth is 4.2-4.5%/yr in all cases.

Why this severity. The very-top rows of Table 1 can reverse (China's top 0.1%+ growing as fast as or faster than Russia's), and the 'more limited' Chinese rise is partly built into the baseline. But the paper's explicit Table 1 claims (China's growth much higher and lifting the bottom 50%; Russia's bottom 50% falling) do not depend on it.

The authors' response. PYZ (2019, fn 19) justify the constant profile 'in the absence of any other information' and by the absence of a trend in observed/simulated tax revenue ratios over 1986-2003 (Piketty & Qian 2009), and present variants in Online Appendix B1.

7moderatelikely · previously known · speculative assumption

The paper's figure that Russians hold offshore wealth worth 85% of national income is not measured. It is built by adding up unexplained gaps in Russia's international accounts since 1992 and assuming that money grew as fast as Russia's economy in dollars. With ordinary investment returns instead, the figure would be about 55-70% for 2015 and 30-45% in earlier years, and the authors' own range of 55-110% is not mentioned. Hidden Russian wealth is clearly large, but the exact number is a rough estimate.

Technical detail and evidence

Why it matters, in detail. Setting r = g makes the offshore/NI ratio equal to the running sum of annual flight/NI ratios (86.9% by 2015). Any fixed portfolio return below Russia's 20-40%/yr dollar NI growth in 2000-08 lowers the pre-2014 path sharply (2013: 32-44% vs 88%), while the 2015 level (56-83%) is inflated by the ruble collapse. NPZ's rationale (offshore holdings reinvested in Russian firms, so they track Russian values) is legitimate, so the direction of bias is not established; the precision of '85%' is.

Evidence. NPZ (2018) p. 205 and fn 23-24; journal ESM1 Table A33 and WP Appendix A Table A33. Re-cumulation (scripts/offshore_sensitivity.py) reproduces the published stocks exactly (journal 86.4% central / 67.2% lower; WP 74.2% / 56.1%). With fixed nominal dollar returns of 3/5/7%, the journal flows give 55.9/67.9/83.1% of NI in 2015 and 32.1/37.5/44.2% in 2013 (vs 88.1% central). The 2015 ratio is propped up by the 2014-15 ruble collapse in the denominator (NI fell from $1,974bn in 2013 to $1,204bn). Russian private wealth excluding offshore is 297% in 2015 (2.4x 1990); with the published lower bounds it is ~2.85-2.95x. The sibling adjudication of NPZ (2018) reports 42-75% under standard alternatives.

Why this severity. This is an undisclosed caveat on a secondary result. The 85% level and the 'gradual increase since 1990' path change materially under conventional return assumptions, and 'more than tripled' becomes ~2.4-3.0x. The existence of large unrecorded wealth, given cumulated trade surpluses of ~230% of NI against small net foreign assets, is not in doubt.

The authors' response. NPZ (2018, p. 205) call the estimates 'by construction relatively conservative', report lower/upper variants (55%/110%), argue that much offshore wealth is reinvested in Russian companies, and note consistency with Zucman's (2014) estimate that about half of Russian financial wealth is held offshore.

8minorconfirmed · newly identified here · arithmetic or transcription

A spreadsheet averaging formula for 2008 left out the crash year 2009, which created the chart's highest point. Russia's lines are also smoothed while the other countries' are not, and readers are not told.

Technical detail and evidence

Why it matters, in detail. Fixing the window lowers the 2008 Russian point by ~10pp of NI, so the series declines 36.0 -> 32.4 -> 31.2% (2007-09) instead of peaking in 2008. Combined with F2, the corrected centred series is 24.4/21.1/19.6% for 2007/08/09. Mixing smoothed and unsmoothed series also invites misleading same-year comparisons (2009: 31.2% plotted vs 12.2% annual).

Evidence. Russia3!BT35 = AVERAGE(BS34:BS35) and BX35 = AVERAGE(BW34:BW35), against e.g. BT36 = AVERAGE(BS35:BS37); BT26 = AVERAGE(BS24:BS27). Plotted 2008 values are 42.5%/40.2%; correctly centred they are 32.4%/30.6%. The annual 2009 Russian ratio is 12.2% (9.2% with a nominal-dollar denominator), whereas the plotted value is 31.2%. The published figure equals the workbook within 0.07pp.

Why this severity. A spreadsheet slip on one plotted point plus a disclosure gap. It changes the shape of the series around 2008 but no conclusion.

9minorconfirmed · newly identified here · arithmetic or transcription

The text says China's housing wealth rose to 182% of national income in 2015, but the paper's own chart shows 201%. The 182% figure, and its 1991 starting year, appear to have been copied from Russia's column. The mistake understates China's housing boom and changes no conclusion.

Technical detail and evidence

Why it matters, in detail. China's 2015 private housing (net of all household debt) is 201% of NI, not 182%: the text understates it by 19pp (38-60pp against the mortgage-net and gross definitions). The rise from 1991 is 140pp rather than 122pp.

Evidence. MainData!I (China, 'housing net debt', plotted in Fig. 2A): 0.61 (1991), 2.01 (2015). PYZ A23 2015: 219.8% net of mortgages, 241.5% gross. MainData!P = Russia3!N (Russia housing net of debt): 60.3% (1991), 182.0% (2015). P&P p. 110.

Why this severity. This is a transcription error in the text, while the correct series is plotted. It understates rather than overstates China's housing boom and changes no conclusion.

10minorconfirmed · newly identified here · version instability

The paper mixes old and updated versions of the Russian wealth figures, so the same 2015 number appears as 383%, 380%, 371% or 347% of national income in different places. The differences are small next to the overall rise, but they make the numbers hard to check.

Technical detail and evidence

Why it matters, in detail. Russian 2015 private wealth is 383% in Figure 1 and the text, 347% in Figure 2B, 371% in the posted file and 367% in the final journal series; offshore wealth is 85% in the text and 74-75% in Figure 2B and the file. The private-wealth multiple 1990-2015 ranges from 2.96x to 3.10x.

Evidence. Digitized published Figure 1 (scripts/digitize_fig1.py) gives Russian private wealth of 383.5% (2015) and 359.4% (2011), up to +18.3pp above MainData!N (370.9% in 2015), and equal within 0.7pp to MainData!N - T + journal ESM1 A33 offshore (383.1% in 2015). China private/public and Russian public lines match the file within 0.7pp. Figure 2B hard-coded components (MainData!W:AA): 1.72 + 0.45 + 0.18 + 0.38 + 0.74 = 347%, i.e. journal A21 housing (1.719), business assets (0.452), equities (0.184) and deposits (0.381) plus the WP offshore stock (0.742). Final journal total: 366.7%. 'More than tripled' holds for the hybrid (3.10x) but not for the file (3.00x) or the journal series (2.96x).

Why this severity. Several vintages of the same quantity (347-383% of NI) appear in one five-page paper. The differences are small relative to the 1990-2015 rise and change no conclusion, but they impede reproduction.

11minorconfirmed · newly identified here · arithmetic or transcription

A handful of numbers in the text are rounded generously or slightly off compared with the authors' own data, for example Russian public wealth of 84% described as 90%. None of these slips changes the story.

Technical detail and evidence

Why it matters, in detail. The text exceeds the data by 5.6pp (Russian public wealth), 6.9pp (Chinese public wealth) and 1.0pp (Russian bottom 50); China's 1980 public share lies 3.1pp below the stated range.

Evidence. Russian public wealth 2015 '90%' vs 84.4% (MainData!U122; 84.8% in the published Figure 1). China public wealth 2015 'approximately 230%' vs 223.1% (MainData!G122). Public share of national wealth 'as large as 70-80 percent in both countries in 1980' vs China 66.9% in 1980 (69.2% in 1978; China3!W). Russian bottom-50 share 2015 'about 18%' vs 17.0% (MainData!AF122).

Why this severity. These are presentation slips within approximate wording; none affects a conclusion.

12minorconfirmed · newly identified here · misrepresentation of source

Russians did not use privatization vouchers to get their flats; tenants were allowed to take ownership of their homes for free, and many did so slowly. The paper gets the mechanism wrong, but its point that housing passed to households gradually is right.

Technical detail and evidence

Why it matters, in detail. The gradual rise of private housing reflects delayed exercise of free tenant-transfer rights (plus revaluation), not voucher redemption; the narrative contrast between housing and enterprise privatization survives.

Evidence. P&P pp. 109-110. NPZ (2018) p. 201: 'housing privatization happened in a more continuous manner than the voucher privatization method used for companies. Tenants were given the right to transfer the housing unit into their ownership free of charge'. World Bank (1995), Russia: Housing Reform and Privatization, Report 14929-RU, executive summary paras 42-45 (checked by Auditor B).

Why this severity. The institutional mechanism is described incorrectly, but the substantive point (gradual transfer of housing vs fast enterprise privatization) and all numbers are unaffected.

13minorconfirmed · newly identified here · version instability

Some table labels give the wrong years: Russia's numbers actually run to 2016 and the US numbers end in 2014. The authors' later data releases also changed several starting and ending values. The effect on growth rates is small, but some comparison levels moved noticeably in later versions.

Technical detail and evidence

Why it matters, in detail. Relabelling Russia as 1980-2016 is the minimal fix. Same-endpoint recomputation raises Russia's rates by 0.02-0.5pp/yr. In the current vintage the 1978/80 China-Russia top-10 gap is 3.3pp (vs 6.1pp) and China's bottom-50 decline is 10.6pp (24.7 -> 14.1) rather than 12.1pp.

Evidence. Russia2!AQ5 = Russia1!B154/Russia1!B118 - 1 and AT5 = (1+AQ5)^(1/36) - 1; Tab1!E4:E11 link to these. Recomputed for the labelled 1980-2015 window: 1.17/-0.55/0.46/3.43/6.42/10.19/13.85/17.52 vs published 1.15/-0.53/0.46/3.34/6.24/9.90/13.44/17.01. PSZ2017MainData 'Data' sheet ends in 2014: 1978-2014 growth is 1.25657% vs Table 1's 1.2565%, and PYZ Data4's US '2015' mean of 65,448 equals the 2014 value x 1.01257. PYZ (2019) Table 2 gives China top 1/0.1/0.01/0.001% of 8.6/9.4/9.5/9.2 vs 8.4/9.1/9.8/10.4. WID bulk data (Oct 2026), sptincj992: Russia 1980 top 10% 24.7% (vs 21.0), bottom 50% 27.7% (vs 31.0), 2015 top 1% 18.6% (vs 20.2); China 1978 bottom 50% 24.7% (vs 27.0). The sibling PYZ adjudication traces the China change to a hard-coded urban/rural ratio (not re-verified here).

Why this severity. Label mismatches shift growth rates by at most ~0.5pp/yr, and later revisions are normal for living series. The revisions narrow the Russia-China 2015 top-1% gap (see F1) and China's 1978-2015 bottom-50 decline, neither of which changes the direction of any finding.

14minorconfirmed · newly identified here · overclaiming

Comparing Soviet and modern Russian living standards is hard because Soviet prices did not reflect shortages and queues. Depending on the method, the poorer half of Russians lost anywhere from about 40% to almost none of their income since 1989. The paper reports only the middle estimate, a 20% fall, and leaves out its source's warning.

Technical detail and evidence

Why it matters, in detail. Bottom-50 income change over 1989-2016 equals g x (16.99/30.13) - 1, which is -39%, -20% or -4% for g = 1.085, 1.41 or 1.70. Russia's Table 1 bottom-50 CAGR (1980-2016) ranges from about -1.3% to ~0%, far below China's 4.5%.

Evidence. NPZ (2018) p. 216 ('very difficult to compare real incomes for the Soviet and post-Soviet periods ... our aggregate growth figures might jump from +41% to +70% or more'; 'little doubt ... that the welfare of the vast majority of the population has improved'). Neef (2022), WIL Technical Note 2022/03 (multipliers 1.3 before 1992, 1.225 in 1992, 1.15 in 1993, 1.075 in 1994). Recomputed (russia_table1_capture.py): bottom-50 income change -20.5% (benchmark), -4.1% at +70% and -38.8% at +8.5%; top-1% capture 56%, 41% and 194% respectively.

Why this severity. The benchmark sits in the middle of a two-sided range, and the sign of the bottom-50 change is negative except at the extreme upper end. The China-Russia contrast in bottom-50 growth survives, so this is an omitted caveat rather than a demonstrated error.

The authors' response. NPZ (2018, p. 216) prefer per-adult national income and acknowledge the comparability problem.

15minorconfirmed · newly identified here · documentation gap

The data file posted with the paper cannot reproduce some key results without other, unpublished spreadsheets. Some numbers shown as data are really assumptions, such as Russia's 1980 public-wealth share, which the source paper never estimated. This makes the paper harder to check but does not by itself change its findings.

Technical detail and evidence

Why it matters, in detail. Readers cannot regenerate Figure 5 or Table 1 from the posted file. The Russian 1980 point in Figure 3 (75%) is assumed rather than estimated (the 1990 estimate is 70.9%), and China's 1980 value (66.9%) lies below the stated '70-80%'.

Evidence. MainData!BD:BU (Figure 5 ratios) are #DIV/0! because their denominator columns are absent; the series exist only in Russia3 and the external RussiaDataWealth sheet. Tab1 China/US values are hard-coded and Russia links to the unpublished NPYZ2018MainFiguresTables.xlsx (externalLink5). China income shares in MainData!AG:AJ are rounded to whole percentage points. MainData!C is labelled 'Housing (net of debt)' but contains gross housing (2.415 in 2015 = PYZ A23 gross). Russia3 row 7 (1980) hard-codes national 4.0, private 1.0 and public 3.0 x NI (public share 0.75), and 1981-89 and 1991-94 are linear interpolations; NPZ (2018, p. 194) 'choose to begin our balance sheet series in 1990' and do not use Soviet estimates. China3!DW45 = DW44 (France 2015) and the UK/Japan/Germany 2014-15 shares repeat 2013. The Figure 3 note cites Piketty & Zucman (2014), but the sheet sources are PSZ 2016 (US), Garbinti-Goupille-Piketty 2016 (France) and WID 23/11/2016 (UK, Japan, Germany). The Figure 4 note describes 'raw estimates' that are not plotted.

Why this severity. These are reproducibility and transparency problems; the underlying country files are public and allowed the exhibits to be rebuilt here. The Russian 1980 public share is an assumption, but a dominant public share in the late Soviet economy is not in doubt (71% in 1990).

16minorlikely · newly identified here · overclaiming

Much of the drop in Russian state wealth in the 1990s happened because Russian companies came to be valued far below their book value, not only because they were handed to private owners. The paper presents the whole drop as the result of voucher privatization.

Technical detail and evidence

Why it matters, in detail. Of the 230pp fall in public wealth, private wealth gained 107pp and market-value national wealth fell 123pp. The residual corporate wealth of 134% of NI (Tobin's Q far below 1) indicates that more than half of the measured decline is a valuation effect, which may itself reflect how privatization was carried out.

Evidence. Russia3, 1990 -> 2000: public wealth S 300.9% -> 70.6%, of which public financial assets (state equity) W 164.1% -> 14.5%; private wealth K 123.8% -> 230.7% (housing N +87pp, largely revaluation); national wealth at market value B 424.7% -> 301.3%; at book value AB 424.7% -> 435.5%; residual corporate wealth (book - market) AH 0 -> 134.2% of NI.

Why this severity. This is an interpretive overstatement of the mechanism; the measured collapse of public property and the shift to private ownership stand.

The authors' response. NPZ (2018, pp. 199-200) note that national wealth fell from over 400% to about 300% of NI between 1990 and 1999.

Claims examined and rejected

Problems one auditor raised that did not survive checking.

In I2, the tax-correction code floors each tail adjustment ratio at one, so the upward correction relative to the raw survey is 'partly imposed by the procedure'. — The floor never binds. In the benchmark the correction ratios at the three knots are 2.05-36.3 in every year 2008-2015 (2.23/8.06/20.54 in 2015; work/russia_taxcorr_years.log), so the upward correction comes entirely from the tax data and the assumed deduction/declaration rates (F1), not from the floor.
In I11, describing Russia's top-1% share as 'approaching' the US level is a text-data mismatch, because Russia was above the US in 2000-2011. — The sentence refers to the period after the 2008-09 fall. In 2012-2015 Russia's top 1% (19.8-21.1%) is at the US level (19.6-20.8%), so the wording is imprecise but not wrong. The substantive problem is that the Russian level itself is assumption-dependent (F1).
In I7, cross-country comparisons of private financial wealth are 'asymmetric by construction' because China receives no offshore imputation. — Narrowed rather than adopted. The P&P (p. 111) explicitly says China's offshore wealth is not estimated and argues it is likely smaller because of tighter capital controls, so the asymmetry is disclosed and plausibly directionally right. The adopted issue (F6) concerns the undisclosed range and the return assumption behind the Russian figure.
I1 is 'major': the Figure 5 conversion error affects a core finding (C5) and its use in the privatization argument. — Severity not adopted. The error is confirmed and large, but Figure 5 is corroborating evidence from which the text draws only the qualitative claim that Russian billionaire wealth is 'much greater', and that claim survives correction (Russia ~2x the US and highest in every year). Graded moderate (F2).
I6 (housing vouchers) is 'moderate' because it misstates the mechanism of the core privatization narrative. — Severity not adopted. The misdescription is confirmed, but it concerns the housing mechanism only. The narrative contrast (gradual housing transfer vs fast enterprise vouchers) and all numbers are unaffected, so it is graded minor (F12).
In I8, the Russian bottom-50 'decline' is within the authors' own uncertainty, and at '+70% or more' its sign flips (graded moderate). — Partly adopted, graded minor (F15). The uncertainty is two-sided: without NPZ's own upward deflator revision, 1989-2016 growth is only ~+8.5% and the bottom-50 fall is ~39%. The benchmark (-20%) is central and the sign is negative except at the extreme upper end of the range.

Audit notes

What reproduced. Reproduced. (i) 30+ numerical text claims against NPYZ2018Data.xlsx: most match; mismatches are listed in F9 and F11. (ii) The published Figures 1 and 5, digitized from the PDF vector paths, match the workbook except Figure 1's Russian private-wealth line, which is a hybrid vintage (F10). (iii) Table 1: the China column matches exactly from PYZ g-percentiles; the Russia column matches exactly with 1980-2016/36 years; the US column equals PSZ 1978-2014. (iv) The growth capture of 56.4% matches NPZ Table 2. (v) Figure 5 numerators match Forbes 2013 exactly. Its denominators were rebuilt from NPZ's own nominal-dollar national income (Appendix A, Table A33) and World Bank NNI, which lowers the 2011-13 Russian values from 37-40% to 21-22% (F2). (vi) NPZ's Stata tax correction was re-implemented in Python, reproducing all nine published variants exactly, and alternative deduction rates were run (F1). (vii) Offshore stocks were re-cumulated exactly and with alternative returns (F6). (viii) China's 1978-baseline sensitivity was computed from the PYZ files (F5). (ix) The current WID vintage was compared for both countries. Not reproduced: the final pretax (copula) step under alternative tax assumptions, for which pretax values are approximated from the fiscal results (consistent with Auditor B's quadrature); national balance sheets from primary statistical sources; the survey microdata and gross-up pipeline; the Forbes 2008 list; and the openICPSR package, which was inaccessible.

Limits of this audit. Focused adjudication of a five-page paper that summarizes PYZ (2017/2019) and NPZ (2017/2018). Pretax top shares under alternative tax assumptions are approximated from re-implemented fiscal shares rather than by re-running the Gpinter/copula step. National balance sheets, survey microdata and the RLMS/HBS processing were not rebuilt from primary sources. The Forbes 2008 list and Hurun-based alternatives for China were not examined. Some inherited points (China's 1978 urban/rural ratio; 2023 Russian tax counts) rely on the sibling adjudications of PYZ (2019) and NPZ (2018) and were not re-verified here. Kapeliushnikov (2020) was read in the HSE preprint version (Russian). The openICPSR deposit (e114470v1) could not be accessed, so the identity of the author-hosted workbook with the deposit is unverified. No author response to the critiques was located. Severity judgments relative to the core findings involve judgement, especially for F1 (major vs moderate) and F2.